Guide · Mutual funds and portfolio construction · Updated 2026-09-19

Every fund house publishes how trapped its small-cap book is, every fortnight, and the number is a floor rather than a forecast

A daily dealing small-cap scheme makes two promises that cannot both be kept on a bad day. It promises that you can leave on any business day at the net asset value struck that evening. It also promises that your money is working in companies whose shares, on that same bad day, barely trade. The first promise is settled in hours. The second takes as long as the order books allow.

The industry's answer to this is not hidden and it is not behind a paywall. Every fund house running a mid-cap or small-cap scheme publishes, in a format identical down to the footnotes, the number of days it would need to sell half that scheme's portfolio. It is republished every fortnight. It is collated in one place. It is the single most informative thing a retail investor can read about a small-cap fund, and almost nobody opens it.

A redemption is priced in a day and funded over weeks

A redemption request received before the cut-off on a business day is priced at that day's net asset value. Net asset value is computed from the closing prices of the securities the scheme holds, each marked at the last price it traded at, whatever quantity that last trade happened to be for. The money leaves the scheme a small number of business days later, on a timeline that tracks the market settlement cycle.

Nothing in that sequence required the scheme to sell anything. The payout can come from cash the scheme was already holding, or from the subscriptions that arrived the same morning. This is why redemptions in ordinary weather are invisible: they net against inflows and the portfolio is never touched. The mismatch is real on every single day and costs nothing on almost all of them.

A redemption wave is the case where they stop netting. Outflows exceed inflows, the cash buffer is drawn down, and the scheme has to raise money by selling. At that point the two clocks come apart. The price was fixed on the day the request arrived. The selling happens over however many days the order books of fifty or eighty small companies will absorb.

The people who stay pay for the people who leave

Net asset value is a valuation, not an execution. The closing price of a thinly traded small company is the price of the last trade of the day, and that last trade may have been a few hundred shares. It is the correct price for valuing the holding. It is not the price at which the scheme could sell its entire position.

So the redeemer is paid at a price derived from the last small trade, and the scheme then has to realise cash from a sale many multiples of that size, into a market its own selling is pushing down. Whatever the selling realises below the price already paid out is absorbed by the units that did not redeem. The transfer is mechanical. Nobody decides it, nobody is at fault for it, and no ordinary rule prevents it.

There is a tool designed for exactly this problem, and it does not apply here. Swing pricing adjusts the net asset value at which dealings are processed during a dislocation, so investors leaving bear the cost of their own exit instead of passing it to those staying. India introduced it for open ended debt schemes with effect from 1 March 2022, excluding overnight and gilt schemes, with the stated intention of examining equity, hybrid and other schemes in later phases. As matters stand it is a debt mechanism. An equity small-cap scheme processes redemptions at the unadjusted net asset value, and the only friction the remaining holders get is the exit load, if the scheme carries one and the units are young enough for it to bite.

The buffer against all of this is cash, and cash is not free: money held against a redemption that has not arrived is money not doing the job the scheme exists to do. The manager is choosing continuously between two costs paid by the same people, which is why the disclosure reports the cash percentage at all.

The disclosure that answers this, published every fortnight, read by almost nobody

The regulatory response to the surge of flows into mid-cap and small-cap schemes was not a restriction on those schemes. It was a disclosure. Fund houses running them were directed to run a prescribed liquidity stress test and publish the results every fifteen days, beginning 15 March 2024, on their own websites and collated by the industry body.

The format is standardised to the footnote. Two schemes at two different fund houses publish the same figures, in the same order, under the same headings, with the same explanatory notes beneath. That is the entire value of the exercise: a liquidity estimate computed by each house to its own method would be worth nothing.

Two things are worth noticing before anything else. The scope is mid-cap and small-cap schemes. And the list mixes liquidity, concentration, volatility and valuation measures, of which only the first two speak to the question the stress test is asking.

How the days figure is built, and why it is a maximum rather than an average

The construction is set out in the notes attached to the format, which is why it can be checked rather than guessed at.

One. Take the portfolio as at a stated date. The figures describe that day's holdings, not today's. The date is printed alongside them for that reason.

Two. Set aside the least liquid fifth. The 20 percent of the portfolio that is hardest to sell, ranked by the liquidity of each security, is removed before the test runs. The stated rationale is that a manager would reasonably hold on to positions taken for the long term rather than dump them into a redemption.

Three. Assume proportional selling. The test assumes the scheme sells every remaining holding in the same ratio as the portfolio composition, and uses its cash in proportion too. This is not a rule any manager has to follow. It is an equal treatment assumption, chosen so that the figure describes what a fair liquidation looks like rather than what a clever one might.

Four. Fix the daily volume available. For each holding, the daily traded volume is taken as the three month daily average across both national exchanges combined, and then multiplied by three. The scheme is assumed able to take a tenth of that without disturbing the price.

Five. Divide, then take the largest. Quantity to be sold divided by quantity sellable per day gives a number of days for each holding. The scheme's published figure is not the average of those numbers. It is the largest of them.

That last step is where the number gets its character. Work a single holding. A scheme has sixty crore in one small company. That company trades fifteen crore a day across both exchanges on a three month daily average. The test grosses that up to forty five crore and allows the scheme a tenth, so four and a half crore a day. Selling half the position, thirty crore, takes a shade under seven trading days. Illustrative figures, prescribed arithmetic.

Now put that holding in a book of fifty names, forty nine of which clear in two days. The published figure is seven, not two, and the forty nine easy positions cannot pull it down. One position sized wrongly against its own order book sets the number for the entire portfolio, which is why the figure moves when a manager builds or unwinds one uncomfortable holding rather than when the market as a whole moves.

It also explains why setting aside the least liquid fifth changes the answer as much as it does. In a maximum, removing the tail does not shave a little off the estimate. It moves the binding constraint to an easier holding, and the printed figure steps down to whatever that one needs.

The volume in the denominator decides the answer

Every liquidity estimate of this shape is a quantity divided by a rate, and the rate is a traded volume. Volume sits in the denominator, so the answer is inversely proportional to whatever volume number is fed in. Double the assumed volume and the days halve. That is not a subtlety of the model. That is the model.

This gives the choice of averaging window far more weight than it appears to carry. The mandated basis is a three month daily average across both exchanges. A three month window that happens to span a quarter of heavy turnover carries that turnover forward into an estimate for a quarter that may look nothing like it. A shorter window tracks the current market more closely and is noisier for exactly that reason, because one large block in a thinly traded company can lift a month in a way it cannot lift a year. A longer window is steadier and can carry the memory of a boom well into a bust. There is no window that is correct. There is only a window that is fixed, and fixing it is what makes two schemes comparable to each other even when neither is comparable to the future.

The important row in that table is not the first. It is the third. At ordinary volume, with no stress assumption whatsoever, the same book takes three times as long as the disclosure prints.

Three assumptions, and two of them push the number down

The volume multiple deserves the most attention because it is doing the most work and it points the wrong way. The case for it is observable: in a sharp fall, turnover often spikes as positions are unwound, so assuming more volume under stress is not absurd on its face. The case against it is the whole mechanism this article is about. Volume spikes in the names everybody is trading. In the bottom half of a small-cap book a fall is not a volume event, it is an absence: the bid thins, the screen widens, and the volume that does appear is the volume of other people trying to do precisely what you are trying to do.

So the published figure is a floor on the difficulty, not a forecast of it. It is worth saying that plainly rather than around it. The number tells you how long a fair, patient, price respecting liquidation would take in a market behaving better than its own three month average. It does not tell you how long a real forced sale would take, and it was never built to.

That does not make it useless, because a floor is informative when the floor is high. A scheme reporting a figure in the tens of days is telling you that even under assumptions chosen to be generous, half its book represents weeks of selling. Whatever the real number turns out to be on the day it matters, it is worse than the printed one.

The two published figures are the same figure

Each scheme publishes two numbers every fortnight: days to liquidate half the portfolio and days to liquidate a quarter of it. Read down any fortnight's file and the same relationship appears row after row. The first is about twice the second.

That is not a coincidence and it says nothing about any scheme. It is a property of the model. Days are a quantity divided by a fixed daily rate, so halving the quantity halves the days, exactly, subject only to rounding to whole days. The quarter figure carries no information the half figure does not already contain.

Which is worth pausing on, because the pair being exactly linear is the clearest available proof that the test does not model market impact. In a real liquidation the second half is harder than the first. The early selling has already absorbed the resting bids, widened the spread and announced to the market what you are doing. A model that captured any part of that would produce a 50 percent figure more than twice the 25 percent figure. These produce exactly twice. The disclosure holds the market still while the scheme sells into it.

Impact cost is the price axis the days number leaves out

There are two ways to be trapped in an illiquid holding. You can take a long time to get out, or you can get out quickly and accept a worse price. The fortnightly disclosure measures the first and is silent on the second, by construction.

Impact cost is the standard Indian measure of the second. The exchanges define it as the percentage degradation against the ideal price when a specified order size is executed, the ideal price being the midpoint of the best bid and the best offer. Two features of that definition carry the whole idea. It is not a property of a security on its own but of a security and an order size together, so quoting an impact cost without saying for what size is meaningless. And it rises faster than order size does, because each further slice of an order consumes a thinner layer of the book than the slice before it.

The stress test's ten percent participation cap is precisely a device for holding impact cost near zero. By refusing to take more than a tenth of a day's volume, the model keeps the selling small relative to the order book and therefore assumes the price does not move against it. The cost of that discipline is time, and time is what the disclosure reports. The trade is real and continuous: a manager under redemption pressure can always have fewer days by paying more concession, and the point on that curve they choose is a judgement the format does not capture and does not claim to.

Impact cost is not a line in the mandated format. If you want the price axis you have to build it, and the inputs are public: the holdings from the monthly portfolio disclosure, and the traded volumes and order book depth published by the exchanges. What the fortnightly file gives you free is the time axis, standardised, for every scheme in two categories, computed on the same date and the same assumptions.

Reading the rest of the format against the days figure

The liquidity figure is the headline and the only one most coverage quotes. The rest of the row is not filler.

The top ten investor share is the demand side of the same equation. The days figure describes what happens if a large redemption arrives. This one describes how concentrated the trigger is, because a book where ten holders own a meaningful slice can lose that slice on a single decision taken in one room.

The market capitalisation split tells you whether the category label is doing what it says. A small-cap scheme must hold a majority in small companies and may place the balance anywhere. A scheme holding a substantial share in large companies has bought itself a liquid sleeve to sell first, which shows up as a better days figure. Worth knowing it came from there rather than from better selection among small companies, because the sleeve can be spent only once.

Cash is the first line of defence, and the test spends it proportionally. The notes state that cash is assumed used on a pro rata basis, which is deliberately not what a manager would do. A real manager spends cash first. The assumption exists to stop a large cash position flattering the liquidity of the equity book underneath it.

Standard deviation and beta are price measures and do not answer this question. A scheme can show a modest standard deviation and carry a poor liquidity profile, because standard deviation measures how much marked prices moved, and marked prices in illiquid securities move less precisely because they trade less often. Reading volatility as a proxy for liquidity gets the sign backwards.

Portfolio turnover is the manager's demonstrated footprint. A high turnover ratio alongside a high days figure describes a manager who trades a great deal in a book that is hard to trade.

Where this gets misread

Reading a low figure as skill. Days scale with the size of the book against the liquidity of the securities in it. A scheme with a small asset base reports fewer days than a large one holding similar companies, and that is arithmetic rather than judgement. The informative reading is the direction of travel for one scheme as its assets grow, not the ranking across schemes of different sizes.

Comparing across categories. A mid-cap figure and a small-cap figure sit in the same file and are not the same measurement, because the underlying liquidity of the two universes differs by a wide margin. Compare within a category or do not compare.

Treating it as a forecast. It is a floor computed on a market assumed to be trading at three times its own average. This misreading survives explanation, because a number printed in whole days looks like a prediction.

Assuming it covers your scheme. The mandate covers mid-cap and small-cap schemes. A thematic or sectoral scheme concentrated in small companies can carry the identical mismatch and publish nothing about it.

Reading a single fortnight. The figure describes a stated portfolio date and moves with both the book and the market. One reading is a point. The series is the information, and the series is free.

What the number is actually for

It is not a sell signal, and a scheme reporting a high figure is not doing anything wrong. A small-cap scheme is supposed to own small companies, and small companies are supposed to be difficult to sell in size. That difficulty is the source of whatever the category is paid for and the source of the risk in the same sentence. The stress test figure is the size of that risk, stated by the manager, in days, on a fixed schedule.

What it is for is position sizing and horizon. A book that needs weeks to exit half of itself is not a place to keep money that might be needed at short notice, and not a place to arrive late in a flow cycle when many other exits are queued behind the same order books. Both are decisions made before investing rather than during a drawdown, which is exactly when the file is easiest to read and hardest to act on.

The wider lesson is that a mandated disclosure is written to a specification, and the specification records both what the regulator wanted known and what it was prepared to leave out. Reading one properly means reading the notes as carefully as the numbers, working out which way each assumption points, and holding the answer at the confidence the method supports rather than the confidence a printed integer implies. That is the same skill whether the document is a stress test file, an annual report or a contract note.